Diego Bonilla

3D-model-From-Single-image. A 3D implementation of the mono-depth neural network for modeling estimation.

Deep-Mouse-Hand-Pose-Estimation. Mouse controller based on hand pose estimation

Believable-Human-Writing-Conversor. A script that transforms any text to human-like writing.

Video_Point_Tracking_OpenCV. Track points in a video using the SIFT algorithm and OpenCV.

IMDB-Database-Automatic-Movie-Picker. Graphical interface to find the perfect movie querying the IMDB database.

Hand-Webcam-Write-AR. Some stuff to be able to twite with one hand

RealTime-3D-Face-Mesh. A better approach following one of my classic repos 🙄

Music-Generator-RNN. AI tries to write music from experience

3D-Reconstruction-From-Two-Views. 3D reconstruction from two images using key points and singular value decomposition.

Full-Model-Fine-tuning-for-FLUX.1-dev-. This project provides a script to perform full model fine-tuning on FLUX.1 [dev]. It is adapted from the original DreamBooth training example in the `diffusers` library.

Time-series-prediction-using-Random-Forest-Regression. Basic times series regression using the Random Forest Regression algorithm

Chain-of-Debate-LLMs. A collaborative AI debate system that leverages multiple AI agents to solve complex problems through structured argumentation and consensus-building.

Human-Like-Typing-Simulator. This project provides a solution for students and educators in environments where typing authenticity is scrutinized

GPT-Image-Corrector. Corrects common visual artifacts in images generated by ChatGPT Image Generator, such as unwanted color casts (e.g., yellow/orange tint) and unnatural sharpness.

Depth-Estimator-From-Single-Image. An inplementation of the monodepth2 model.

Structure-from-motion-RGBd-Camera. Structure from motion using Intel Realsense Depth camera

ConvMixers-Auto-Encoder. An idea for a generator network using the new ConvMixer's architecture.

Audio-Speech-Noise-Removal. Another nice Unet project

Chess-Game-AI-MINIMAX. This. Is. An. Algorithm. Not. An. IA.

RNN-MIDI-Music-Generator. A better version of my old Music Generator.

Most-Interesting-Deep-Learning-Papers-Compilation. Just some compilation I've made of the papers I found most interesting.

Bit-Depth-Upscaler. BitDepthUpscaler: Safe Residual Dequantization Networks for 8-bit to 16-bit Image Enhancement

SketchRecognition-Sketch-to-Face-Retrieval-and-Reconstruction. SketchRecognition is a sketch-to-face recognition and reconstruction system that bridges forensic sketches with standard face recognition pipelines. It trains a dedicated sketch encoder that maps sketches into the ArcFace embedding space while keeping ArcFace unchanged.

3D-AR-PingPong. 3D AR PingPong using various types of detection algorithms.

Homeostatic-Neural-Networks. Non-official and toy implementation of the fun paper: "Need is All You Need: Homeostatic Neural Networks Adapt to Concept Shift"

Gemini-Video-Highlights. Gemini creates a short highlights video compilation of a given long video. This idea leverages Google's Gemini LLM, specifically its long attention and caching capabilities, to handle extensive video inputs effectively.

MTG-Card-Creation. MTG Card Creator: A Fusion of GPT and Stable Diffusion

ViT-Assembler. This repository presents a self-supervised learning approach utilizing a Vision Transformer (ViT) Encoder to solve a jigsaw puzzle as a pretext task.

CERMatch. CERMatch is a novel Python library designed for evaluating Optical Character Recognition (OCR) systems using Character Error Rate (CER) based metrics. This library provides a unique method for matching ground truth text words with predicted words, offering a comprehensive analysis of OCR accuracy.

Vision-Transformers-Pytorch-Implementation. Python

Gemini-Becomes-an-Expert. By converting user queries into targeted searches across multiple platforms the algorithm ensures a thorough collection of information.

Drawing-Assistant-CLIP. Using CLIP to guess whatever you can imagine drawing.

Unsupervised-Domain-Adaptation-Pytorch. Semipersonal implementation of the paper "Unsupervised Domain Adaptation by Backpropagation"

AdaIN-StyleGAN-Keras. My implementation of the paper "Arbitrary Style Transfer in Real-time with Adaptive Instance Normalization"

BEGAN-keras. A BEGAN implementation using idiomatic keras

DL-Auto-Green-Screen. Auto person matting and color adjustment

Image-To-High-Resolution-With-Cycle-GAN. Image to high resolution using just one python library.

Augmented-Reality-3D-Model-Scene-Placement. A PokemonGO look-a-like marker based Augmented Reality demo.

3D-Face-Projection-from-landmarks. Some script using the 3D MatPlotLib and the dlib face landmarks detection.

Google-Cloud-Vision-AI-Test1. First script using the Google Vision AI API.

Detectron2-Facebook-COCO-dataset. Easy CPU implementation of the Facebook's Detectron2 object detector.

Hand-Posture-Recognition-OpenCV. Hand posture recognition using skin detection and structural analysis in OpenCV

Tensorflow-SSD-Object-Detection. Object Location using Mobile Neural Network and Convolutional Neural Network